Regression methods for binomial and Poisson distributed data

Conference · Sun Jan 01 00:00:00 EST 1984

OSTI ID:5003503

Frome, E L

Models are considered in which a rate or probability can be represented by a regression function that describes the relation between the predictor variable and the unknown parameters. Estimates of the parameters can be obtained by means iteratively reweighted least square (IRLS). When the dependent variable is a count that follows either the Poisson or binomial distribution, the IRLS algorithm is equivalent to using the method of scoring to obtain maximum likelihood (ML) estimates. This general least squares regression approach includes linear, generalized linear, and intrinsically nonlinear regression functions. Standard statistical packages that support IRLS can be used to obtain ML estimates, their asymptotic covariance matrix, and diagnostic measures that can be used to aid the analyst in detecting outlying responses and extreme points in the model space. The results of fitting several different models to the same data set can be summarized in an ANOVA-like table using the deviance as a measure of residual variation. Five examples using data from both designed experiments and observational studies are presented to illustrate the utility of Poisson and binomial regression analysis.

• Research Organization: Oak Ridge National Lab., TN (USA)
• DOE Contract Number: AC05-84OR21400
• OSTI ID: 5003503
• Report Number(s): CONF-8403102-1; ON: DE84009714
• Resource Relation: Conference: 1. midyear topical symposium of American Association of Physicist in Medicine, Mobile, AL, USA, 14 Mar 1984; Other Information: Portions are illegible in microfiche products
• Country of Publication: United States
• Language: English

Related Subjects

71 CLASSICAL AND QUANTUM MECHANICS
GENERAL PHYSICS
REGRESSION ANALYSIS
POISSON EQUATION
ITERATIVE METHODS
LEAST SQUARE FIT
MATHEMATICAL MODELS
STATISTICS
DIFFERENTIAL EQUATIONS
EQUATIONS
MATHEMATICS
MAXIMUM-LIKELIHOOD FIT
NUMERICAL SOLUTION
PARTIAL DIFFERENTIAL EQUATIONS
657006* - Theoretical Physics- Statistical Physics & Thermodynamics- (-1987)
658000 - Mathematical Physics- (-1987)